rakki-18 / matrix-benchmarks Goto Github PK

When the user clicks on the benchmark button, without selecting an operation to benchmark, an error is obtained.

Steps to reproduce

• MBApplication new start to run the benchmark UI.
• Click on the benchmark button right away.

Add a double-click action or contextual menu with an item to open a Browser on the selected operation in the operations list. The standard way to implement it now in Pharo 9 uses the Commander 2 framework.

This would require to model or store a benchmark result object (with timestamps) into a Collection of results. These results could be available through a UI list which contains the time and the benchmark which was ran.

It could be integrated into the current Results tab, or in another tab such as Previous executions.

Note that from the current implementation it is not possible to determine which benchmark was ran just looking in the Results tab (so one has to remember which operation was the last one selected).

It's difficult to find out in which tab I am situated. For example Single Benchmark or Multiple Benchmark.
It could be nice to highlight the "active" tab representing the selected page in the notebook presenter.

The Results tab has much space available which could be used by adding another presenter for the Roassal canvas. This canvas would include the stacked histograms from the results of the last executions.

For the warning message when clicking the Plot tab without running any benchmark first, it's commonly used an alert message like this:

UIManager default alert: 'You need to run a benchmark first' title: 'Warning'.

The results tab seems to be lost after selection of another benchmark.
It would be nice to have a button to export the results into CSV format (using NeoCSVWriter).

How to reproduce:

1. Run "Accessing a cell in a matrix" Single benchmarks.
2. Access the Plots tab.
3. Go back to Single benchmarks tab.
4. Select "Matrix multiplication" benchmark item.
5. Run it.
6. Plots tab are not updated with new benchmark results plots.

It would be nice to have a progress bar once the start buttons was hit.
This can be do by implementing #do:displayingProgress in the main iteration loop.

There are a few matrix implementations in Libtensorflow and Smallapack.

These have to be included along with the other implementations in the benchmark.